Merge pull request #77 from jasonrandrews/master

use sse2neon for aarch64 performance improvements

.gitlab-ci.yml

@@ -51,9 +51,6 @@ build_ubuntu_bionic:
     - *prepare_artfacts
   script:
     - bash builder/build-package ubuntu bionic
-  only:
-    variables:
-      - $CI_COMMIT_MESSAGE =~ /\[full [Cc][Ii]\]/
   artifacts:
     paths:
       - output/
@@ -70,9 +67,6 @@ build_ubuntu_bionic_arm:
     - *prepare_artfacts
   script:
     - bash builder/build-package ubuntu bionic
-  only:
-    variables:
-      - $CI_COMMIT_MESSAGE =~ /\[full [Cc][Ii]\]/
   artifacts:
     paths:
       - output/
@@ -86,7 +80,7 @@ build_ubuntu_bionic_libjpeg_turbo:
   after_script:
     - *prepare_artfacts
   script:
-    - bash builder/build-package ubuntu focal +libjpeg-turbo_latest
+    - bash builder/build-package ubuntu bionic +libjpeg-turbo_latest
   artifacts:
     paths:
       - output/
@@ -120,6 +114,9 @@ build_ubuntu_focal_arm:
     - *prepare_artfacts
   script:
     - bash builder/build-package ubuntu focal;
+  only:
+    variables:
+      - $CI_COMMIT_MESSAGE =~ /\[full [Cc][Ii]\]/
   artifacts:
     paths:
       - output/

CMakeLists.txt

@@ -210,7 +210,13 @@ endif()
 set(HAVE_PAM ${ENABLE_PAM})
 
 # Check for SSE2
-check_cxx_compiler_flag(-msse2 COMPILER_SUPPORTS_SSE2)
+# Arm is not SSE2 but say it is and use sse2neon.h to convert to neon
+check_cxx_compiler_flag("-march=armv8-a" COMPILER_ARM)
+if(COMPILER_ARM)
+  set(COMPILER_SUPPORTS_SSE2 1)
+else()
+  check_cxx_compiler_flag(-msse2 COMPILER_SUPPORTS_SSE2)
+endif()
 
 # Generate config.h and make sure the source finds it
 configure_file(config.h.in config.h)

builder/README.md

@@ -135,6 +135,7 @@ These instructions assume KasmVNC has been cloned at $HOME and ```kasm_www.tar.g
 cd ~
 tar -zxf kasm_www.tar.gz -C KasmVNC/builder/
 cd KasmVNC
+sed -i 's/^build_www_dir$/#build_www_dir/' builder/build-tarball
 sudo builder/build-package ubuntu bionic
 ```
 The resulting deb package can be found under ~/KasmVNC/builder/build/bionic

builder/dockerfile.ubuntu_bionic+libjpeg-turbo_latest.build

@@ -1,15 +1,14 @@
-FROM ubuntu:20.04
+FROM ubuntu:18.04
 
 ENV KASMVNC_BUILD_OS ubuntu
-ENV KASMVNC_BUILD_OS_CODENAME focal
-ENV XORG_VER 1.20.8
+ENV KASMVNC_BUILD_OS_CODENAME bionic
+ENV XORG_VER 1.20.10
 
 RUN sed -i 's$# deb-src$deb-src$' /etc/apt/sources.list
 
 RUN apt-get update && \
       apt-get -y install sudo
 
-RUN DEBIAN_FRONTEND=noninteractive apt-get install -y --no-install-recommends tzdata
 RUN apt-get update && apt-get -y build-dep xorg-server libxfont-dev
 RUN apt-get update && apt-get -y install cmake git libgnutls28-dev vim wget tightvncserver
 RUN apt-get update && apt-get -y install libpng-dev libtiff-dev libgif-dev libavcodec-dev libssl-dev
@@ -26,6 +25,12 @@ RUN cd /tmp/libwebp-1.0.2 && \
     ./configure --enable-static --disable-shared && \
     make && make install
 
+# Fix for older required libs
+#RUN cd /tmp && wget http://launchpadlibrarian.net/347526424/libxfont1-dev_1.5.2-4ubuntu2_amd64.deb && \
+#    wget http://launchpadlibrarian.net/347526425/libxfont1_1.5.2-4ubuntu2_amd64.deb && \
+#    dpkg -i libxfont1_1.5.2-4ubuntu2_amd64.deb && \
+#    dpkg -i libxfont1-dev_1.5.2-4ubuntu2_amd64.deb
+
 RUN useradd -m docker && echo "docker:docker" | chpasswd && adduser docker sudo
 
 COPY --chown=docker:docker . /src

builder/dockerfile.ubuntu_bionic+libjpeg-turbo_latest.deb.build

@@ -1,7 +1,5 @@
-FROM ubuntu:20.04
+FROM ubuntu:bionic
 
-RUN apt-get update && \
-      DEBIAN_FRONTEND=noninteractive apt-get install -y --no-install-recommends tzdata
 RUN apt-get update && \
       apt-get -y install vim build-essential devscripts equivs
 

builder/dockerfile.ubuntu_bionic+libjpeg-turbo_latest.deb.test

@@ -1,4 +1,4 @@
-FROM ubuntu:20.04
+FROM ubuntu:bionic
 
 ENV DISPLAY=:1 \
     VNC_PORT=8443 \

common/rfb/CMakeLists.txt

@@ -107,7 +107,12 @@ set(SCALE_DUMMY_SOURCES
   scale_dummy.cxx)
 
 if(COMPILER_SUPPORTS_SSE2)
-  set_source_files_properties(${SSE2_SOURCES} PROPERTIES COMPILE_FLAGS ${COMPILE_FLAGS} -msse2)
+  if(COMPILER_ARM)
+    # This is for Graviton2, adjust for other CPUs: -march=armv8-a+crc+crypto
+    set_source_files_properties(${SSE2_SOURCES} PROPERTIES COMPILE_FLAGS ${COMPILE_FLAGS} -march=armv8.2-a+fp16+rcpc+dotprod+crypto)
+  else()
+    set_source_files_properties(${SSE2_SOURCES} PROPERTIES COMPILE_FLAGS ${COMPILE_FLAGS} -msse2)
+  endif()
   set(RFB_SOURCES
     ${RFB_SOURCES}
     ${SSE2_SOURCES}

common/rfb/cpuid.cxx

@@ -51,7 +51,9 @@ namespace rfb {
 
 bool supportsSSE2() {
 	getcpuid();
-#if defined(__x86_64__) || defined(__i386__)
+#if defined(__aarch64__)
+        return true;
+#elif defined(__x86_64__) || defined(__i386__)
 	#define bit_SSE2        (1 << 26)
 	return cpuid[3] & bit_SSE2;
 #endif

common/rfb/scale_sse2.cxx

@@ -16,7 +16,11 @@
  * USA.
  */
 
+#ifdef __aarch64__
+#include "sse2neon.h"
+#else
 #include <emmintrin.h>
+#endif
 
 #include <rfb/scale_sse2.h>